Emission Allowances in the European Union Emissions Trading System

Sammanfattning: The first part of the thesis analyses the short term behavior of daily emission allowance (EUA) log returns with a focus on volatility dynamics in the recent market environment. In this part, I present a historical overview of the European Union Emission Trading System (EU ETS), analyze the stylized facts of the time series, employ appropriate time series models, and assess model in-sample and out-of-sample performance. Due to the existence of leptokurtosis and volatility clustering in the time series, I implement three GARCH models. In addition to a simple GARCH model, I analyze a GJR-GARCH model and an EGARCH model to assess the existence of a leverage component. The performance of the three models is examined and benchmarked against a naive model, which is not incorporating any conditional variance modeling. I examine the models' performance by conducting in-sample goodness of fit and out-of-sample forecasting analysis. The findings strongly support the appropriateness of models capturing leptokurtosis and volatility clustering in the time series while not unambiguously confirming the suitability of models incorporating a leverage effect. The second part of the thesis analyses the long-run relationship between emission allowances (EUA) and European natural gas (TTF). I determine whether emission allowances and natural gas have a long-run relationship, whether this relationship has changed over the considered period, and which market leads the price discovery process. The hypothesis I suggest in support of the existence of a long-run relationship between the EUA and TTF time series is that fuel switching from coal to natural gas is one of the main practically achievable and available options to power producers in order to abate emissions and that the fuel switching decision is connected to the price of emission allowances. The empirical results suggest that there is a cointegration relationship between the two time series in the sample period covering the calendar year 2019 while showing no cointegration relation in earlier sample periods. As a logical consequence, one can assume that the relationship between the two variables has changed over time. The estimated parameter coefficients of the Vector Error-Correction Model (VECM) are employed to analyze the price discovery process. The results do not unequivocally support the assumption that EUAs drive the price discovery process.